Observed Changes and Projected Risks of Hot–Dry/Hot–Wet Compound Events in China

Abstract

Compound extreme events can cause serious impacts on both the natural environment and human beings. This work aimed to explore the changes in compound drought–heatwave and heatwave–extreme precipitation events (i.e., CDHEs and CHPEs) across China using daily-scale gauge-based meteorological observations, and to examine their future projections and potential risks using the Coupled Model Intercomparison Project (CMIP6) under the shared socioeconomic pathway (SSP) scenarios (i.e., SSP1-2.6, SSP2-4.5, and SSP5-8.5). The results show the following: (1) The frequencies of CDHEs and CHPEs across China showed a significant increasing trend from 1961 to 2020, with contrasting trends between the first half and second half of the period (i.e., a decrease from 1961 to 1990 and an increase from 1991 to 2020). Similar trends were observed for four intensity levels (i.e., mild, moderate, severe, and extreme) of CDHEs and CHPEs. (2) All the frequencies under three SSP scenarios will show increasing trends, especially under higher emission scenarios. Moreover, the projected intensities of CDHEs and CHPEs will gradually increase, especially for higher levels. (3) The exposure of the population (POP) and Gross Domestic Product (GDP) will be concentrated mainly in China’s coastal areas. The GDP exposures to the CDHEs and CHPEs will reach their highest values for SSP5-8.5, while the POP exposure will peak for SSP2-4.5 and SSP5-8.5, respectively. Our findings can offer scientific and technological support to actively mitigate future climate change risks.

1. Introduction

The Intergovernmental Panel on Climate Change (IPCC) has reported that Earth’s surface temperature mean in 2013–2022 increased by approximately 1.15 °C compared with the average temperature in 1850–1900 [1,2]. This increase in temperature has accelerated the velocity of hydrological cycles across global areas. Reinforced by human activities, this phenomenon has resulted in an elevated frequency of extreme weather and climate events, such as droughts, heatwaves, extreme precipitation, and floods [3,4,5]. However, these extreme events often do not occur singularly. Interacting physical processes under changing environments often cause various extreme events to occur simultaneously, known as compound events [6]. Compared with individual events, compound events tend to be more dangerous and significantly impact humans and the natural environment [7]. Therefore, it is important to understand how compound events are changing across different spatial and temporal scales, particularly in vulnerable areas (e.g., China, Europe).

As stated by the special report from the IPCC in 2012 [8], various definitions of compound events were proposed [6,9], with more attention given to temperature–precipitation compound events in the context of global warming [10,11], including compound drought–heatwave events (CDHEs) [12,13] and compound heatwave–precipitation events (CHPEs) [14,15,16]. Thus, many studies focused on the changes in CDHEs and CHPEs at different scales. For example, the occurrence probability and intensity of CDHEs have been increasing globally in many regions since the 1950s [1], particularly in drylands [13] and the tropics [17,18]. Examples of typical CDHEs occurred in Europe (2003, 2018), Russia (2010), California in the United States (2014), and China (2022) [19,20,21,22], resulting in devastating impacts on humans and economic systems. Similarly, there has been a significant increase in CHPEs in most regions of the world over the last three decades [23,24]. The occurrence of CHPEs has become more frequent globally in recent years, with notable examples being the United States in September 2017, Japan in July 2018, the United Kingdom in August 2020, and South Korea in July 2020 [23,25,26], leading to severe infrastructure damage and livestock deaths in each case.

Additionally, many studies were also conducted to explore the projected changes in the CDHEs and CHPEs, as well as their possible risks due to the enormous impacts [15,27]. Many regions are projected to experience an increase in the frequencies and intensities of CDHEs/CHPEs with higher global warming [27,28], particularly under high-emission scenarios [29,30]. For example, Bevacqua et al. [12] highlighted that there will be stronger increases in the occurrence of CDHEs with global warming, but this is associated with large uncertainties; De Luca and Donat [31] also confirmed that CDHEs are projected to increase over large parts of the globe by the end of the 21st century; Wu et al. [32] found that CDHEs mainly occur in Europe, South Africa, and the Amazon, while CHPEs mainly occur in the eastern USA, eastern and southern Asia, Australia, and central Africa; Zhou et al. [33] showed that the frequency and intensity of CHPEs are projected to increase fourfold over half of the global land regions under high-emission scenarios. However, uncertainty in the future due to different Global Climate Models (GCMs) or emission scenarios causes inconsistent or opposite trends in CDHEs/CHPEs for the same regions and same periods [34,35,36,37]. Compared with the Coupled Model Intercomparison Project phase 5 (CMIP5), climate projections of GCM simulations from the CMIP6 can more accurately identify spatiotemporal variations in temperature extremes [38,39], heatwaves [40], and extreme precipitation [41,42] due to CMIP6 combining Shared Socioeconomic Pathways (SSPs) and Representative Concentration Pathways (RCPs) [43,44]. Furthermore, the Multi-Model Ensemble (MME) of CMIP6 was shown to mitigate the biases of individual models, leading to more accurate results [37,45,46,47]. As a result, future climate projections based on CMIP6-MME have become widely utilized for assessing climate change risks.

China’s complex and diverse climate is due to its unique geographical location and topographical characteristics, making it sensitive to climate extremes, particularly under global warming and rapid urbanization. Recently, various extreme events have occurred in China, such as extreme rainfall over Henan Province in 2021 [48], extreme high-temperature events across the Yangtze River in 2022 [10], and extreme floods in Haihe River in 2023. Certainly, compound extreme events are also frequent in China. Thus, many studies paid more attention to this issue in recent years, where they focused on historical changes, future projections, possible causes, and impacts at different scales [49,50,51,52,53]. Previous studies indicated an increasing trend in the frequencies and intensities of both CDHEs and CHPEs for most regions in China but with obvious differences in magnitudes due to large uncertainties [15,23,54,55,56,57,58]. However, only a limited number of studies investigated the changes in both CDHEs and CHPEs simultaneously. Thus, it is crucial to prioritize studying the varying types of compound extreme events to prevent and mitigate natural disasters and ensure high-quality economic development.

In this study, we comprehensively investigated the historical variations in both compound events throughout China concerning projected future risks under different scenarios, as well as the changes in socio-economic exposures under the two types of compound events. The main research objectives were as follows: (1) Characterize the temporal and spatial changes in the frequencies of CDHEs and CHPEs under historical and different future scenarios. (2) Construct magnitude indices to examine the CDHE and CHPE intensity variations at four levels. (3) Analyze the socio-economic exposure of CDHEs and CHPEs under different future scenarios.

2. Data and Methods

2.1. Study Area

China has a complex and diverse climate that encompasses frigid, temperate, and tropical zones. It is significantly affected by climate extremes and located near the northwest Pacific Ocean in East Asia. As a result, it has been a focal point in climate change studies. It is also one of the countries with the largest number of rivers in the world. The number of rivers with a catchment area larger than 1000 km² has reached more than 1500 [59]. China’s rivers can be roughly divided into nine major river basins: the Pearl River Basin (PRB), Yangtze River Basin (YZRB), Southwest River Basin (SWRB), Inland River Basin (IRB), Songhua and Liaohe River Basin (SLRB), Yellow River Basin (YRB), Huai River Basin (HURB), Hai River Basin (HRB), and Southeast River Basin (SERB), as shown in Figure 1. Boundary data for the nine watersheds were derived from the Resource and Environmental Science Data Platform (RESDC) https://www.resdc.cn/ (accessed on 7 August 2023).

2.2. Data

2.2.1. Meteorological Data

The gauge-based daily precipitation and temperature data used in this study were obtained from the China Meteorological Administration (CMA) https://data.cma.cn/ (accessed on 5 July 2023). These data were subjected to strict quality control to ensure spatial consistency and continuity [60], and were widely used in previous studies [11,49,50,51]. Here, the criteria for selecting the meteorological stations included (1) the period of data collection should be from 1 January 1961 to 31 December 2020; (2) the amount of missing data from each station should not exceed 10% of the current year’s data; and (3) the percentage of missing values in the full period should not exceed 1% at any station. Thus, a total of 1768 stations were finally screened, as shown in Figure 1. Moreover, in this work, we interpolated these observed records from the filtered meteorological stations into raster data with a spatial resolution of 0.2° × 0.2° using the Inverse Distance Weighting (IDW) method [61,62].

2.2.2. GCMs Data

GCMs data were obtained from the World Climate Research Programme (WCRP) CMIP6 [43,44] https://esgf-node.llnl.gov/search/cmip6/ (accessed on 10 August 2023) and screened based on the following criteria: (1) both the historical and projection data based on various scenarios (e.g., the low-radiative-forcing scenario (SSP1-2.6), the medium-radiative-forcing scenario (SSP2-4.5), and the high-radiative-forcing scenario (SSP5-8.5)) were available at a daily scale, with the variables including the precipitation, near-surface air temperature, and maximum near-surface air temperature; (2) the control indices of the CMIP6 models were all selected as “r1i1p1f1”, where the r, i, p, and f are the realization index, initialization index, physics index, and forcing index, respectively; (3) to ensure data diversity, only one model with the higher data quality applied was selected from each institution in China. Thus, twelve CMIP6 models were finally used in this work, as shown in

Table 1. Details of the CMIP6 models used in this work.

No.	Model	Country	Institution	Resolution
1	ACCESS-CM2	Australia	CSIRO-BOM	1.88° × 1.25°
2	CanESM5	Canada	CCCMA	2.81° × 2.79°
3	CMCC-ESM2	Italian	CMCC	1.25° × 0.94°
4	EC-Earth3	Europe	EC-Earth-Cons	0.7° × 0.7°
5	FGOALS-g3	China	CAS	2.0° × 2.28°
6	GFDL-ESM4	USA	GFDL	1.25° × 1°
7	IPSL-CM6A-LR	France	IPSL	2.5° × 1.27°
8	MIROC6	Japan	JAMSTEC	1.41° × 1.4°
9	MRI-ESM2-0	Japan	MRI	1.13° × 1.12°
10	NESM3	China	NUIST	1.88° × 1.86°
11	NorESM2-LM	Norway	NCC	2.5° × 1.89°
12	TaiESM1	China	AS-RCEC	1.25° × 0.94°

. First, given the different spatial resolutions for these models, we used bilinear interpolation [11] to convert the resolutions of the three variables to a 0.5° × 0.5° grid. Subsequently, the MME of the twelve models was computed to facilitate a bias correction.

2.2.3. POP and GDP Data

In this study, the POP and GDP data were derived from the gridded database of the population and economy under various SSP scenarios, provided by Jiang et al. [63]. This dataset, founded on the Sixth National Population Census of China and the Shared Socioeconomic Pathways, employs two models—the Population–Development–Environment (PDE) and Cobb–Douglas models—to calculate the population and economic data from 2010 to 2100. It further incorporates variations in the POP and GDP under different fertility policies [64], aligning more accurately with actual changes in China. This study selected the POP and GDP data under the SSP1, SSP2, and SSP5 scenarios from 2015 to 2100, with a resolution of 0.5° × 0.5°. In this study, the POP was measured in millions of people, and the GDP was measured in billions of United States dollars (USD).

2.3. Methods

2.3.1. Definitions of Drought, Heatwave, and Extreme Precipitation

First, a drought event was identified using the Standardized Precipitation Evapotranspiration Index (SPEI), with a threshold of one-month SPEI (i.e., SPEI-1) set at −0.8 [51,65]. Compared with the Standardized Precipitation Index (SPI), the SPEI has a wider range of applications and more accurate results [66,67]. Here, we used the Thornthwaite method to calculate the potential evapotranspiration [68], and then estimated SPEI based on the procedure proposed by Vicente-Serrano et al. [69] and Wang et al. [70]. Moreover, a heatwave event was defined as three or more consecutive days (i.e., ≥3 days) with maximum temperatures above 30 °C and the 90th percentile threshold [49]. Similarly, an extreme precipitation event was defined as a day when the daily precipitation exceeded the 90th percentile threshold of the daily precipitation series [71].

Subsequently, we developed magnitude indexes to measure the intensities of three events (i.e., drought, heatwave, and extreme precipitation), characterized by the Drought Magnitude Index (DMI), Heatwave Magnitude Index (HMI), and Precipitation Magnitude Index (PMI) [51,56,65].

(1)

Construction of the DMI

When a drought event occurred in the $i$th month, the difference (ΔD_i) between the SPEI and the drought threshold was estimated (Equation (1)), and then normalized to obtain the DMI_i (Equation (2)) [51,65]:

$$\Delta D_i=\left|D_i-\left(-0.8\right)\right|$$

(1)

$${DMI}_i=0.9\times {\displaystyle \frac{\Delta D_i-\Delta D_{min}}{\Delta D_{max}-\Delta D_{min}}}+0.1$$

(2)

where $D_i$ represents the SPEI value for the $i$th month, while $\Delta D_{min}$ and $\Delta D_{max}$ represent the minimum and maximum values of $\Delta D_i$ in the series, respectively.

(2)

Construction of the HMI

In a heatwave event, the difference (ΔT_i) between the daily maximum temperature and the heatwave threshold was calculated (Equation (3)) and then normalized to obtain $R\Delta T$ (Equation (4)). The $R\Delta T$ during the period of a given drought event was summed to obtain the HMI for this heatwave event (Equation (5)) [51,56].

$$\Delta T_i=T_i-T_{threshold}$$

(3)

$$R\Delta T_i=0.9\times {\displaystyle \frac{\Delta T_i-\Delta T_{min}}{\Delta T_{max}-\Delta T_{min}}}+0.1$$

(4)

$$HMI=\sum _{i\ge 3}^{H_i}R\Delta T$$

(5)

where $T_i$ represents the highest temperature on the $i$th day of the heatwave event; $T_{threshold}$ represents the threshold of a heatwave event; $\Delta T_{max}$ and $\Delta T_{min}$ represent the maximum and minimum values of $\Delta T$ during a single heatwave event; and $H_i$ represents the duration of the heatwave.

(3)

Construction of the PMI

Similar to the HMI, for a single extreme precipitation event, the difference ($\Delta P)$ between the extreme precipitation and the threshold was first calculated (Equation (6)). Then, the $\Delta P$ was normalized (Equation (7)). Generally, extreme precipitation events may be discontinuous in a certain period. Finally, the normalized $R\Delta P$ of extreme precipitation during a certain period was summed up to obtain the PMI (Equation (8)) [56].

$$\Delta P_i=P_i-P_{90\mathrm{th}}$$

(6)

$$R\Delta P_i=0.9\times {\displaystyle \frac{\Delta P_i-\Delta P_{min}}{\Delta P_{max}-\Delta P_{min}}}+0.1$$

(7)

$${PMI}_i=\sum _{i\ge 1}^{P_j}R\Delta P$$

(8)

where $P_i$ represents the precipitation volume on the $i$th day of the extreme precipitation event, while $P_{90\mathrm{th}}$ represents the threshold value of the 90th percentile of the precipitation volume. $\Delta P_{max}$ and $\Delta P_{min}$ represent the maximum and minimum values of $\Delta P$ within a single extreme precipitation event, respectively. $P_j$ represents the duration of the extreme precipitation.

2.3.2. Definition of CDHE and CHPE

In this work, the CDHEs and CHPEs are defined as shown in Figure 2. A CDHE is defined as drought and heatwave events occurring in a given month simultaneously. If two heatwaves occur during the month, they are counted as two CDHE events. The formula for the Compound Drought–Heatwave Magnitude Index (CDHMI) is shown in (Equation (9)) [51].

$$CDHMI=DMI\times HMI=R\Delta D\times \sum _{i\ge 3}^{H_i}R\Delta T$$

(9)

A CHPE is defined as a heatwave event and an extreme precipitation event occurring in a given month during the heatwave. If multiple extreme precipitation events occur, it is still treated as a CHPE. The formula for the Compound Heatwave–Extreme Precipitation Magnitude Index (CHPMI) is shown in (Equation (10)) [56].

$$CHPMI=HMI\times PMI=\sum _{i\ge 3}^{H_i}R\Delta T\times \sum _{i\ge 1}^{P_j}R\Delta P$$

(10)

To have a more intuitive response to the intensity of CDHEs and CHPEs, the cumulative probability density distribution (PDF) curves were used to classify the intensities of the CDHMI and CHPMI. Referring to previous research [51,56,65], the magnitude indices for 1961–2020 were divided into four levels: mild (0–50%], moderate (50–75%], severe (75–95%], and extreme (95–100%], as shown in

Table 2. Classifications of CDHMI and CHPMI.

	Mild	Moderate	Severe	Extreme
CDHMI	(0, 1.19]	(1.19, 2.36]	(2.36, 5.8]	>5.8
CHPMI	(0, 2.56]	(2.56, 4.38]	(4.38, 10.36]	>10.36

.

2.3.3. Delta Downscaling and Simulation Assessment

It is challenging to use GCMs for regional studies due to their lower resolution. Delta downscaling is widely used in climate studies to reduce the model-measured error and accurately reflect future climate changes [72,73]. Delta downscaling has shown excellent results in China [74,75]. Thus, CMIP6 data were bias-corrected using Delta downscaling in this study. Then, the grid CMIP6-MME data for the three future scenarios were bias-corrected by combining the observed meteorological data with the CMIP6 historical data. The precipitation was calculated using Equation (11), and the mean and maximum temperatures were calculated using Equation (12) [72,73].

$$P_{delta}={\displaystyle \frac{P_{mme-fut}}{P_{mme-his}}}\times P_{obs}$$

(11)

$$T_{delta}=\left(T_{mme-fut}-T_{mme-his}\right)+T_{obs}$$

(12)

where $P_{obs}$ and $T_{obs}$ represent the observed precipitation and temperature data during 1961–2014; $P_{mme-fut}$ and $T_{mme-fut}$ represent the projected CMIP6-MME precipitation and temperature data from 2025 to 2099; and $P_{mme-his}$ and $T_{mme-his}$ represent the historical CMIP6-MME precipitation and temperature data from 1961 to 2014.

In this study, each GCM was evaluated using Taylor diagrams. The Correlation Co-efficients (CCs), Root Mean Square Deviations (RMSDs), and Standard Deviations (SDs) were calculated for the 12 CMIP6 models, MME, and MME-Delta, respectively. The formulas for these three indicators are given by Ma et al. [47].

$$SD={\displaystyle \frac{\sqrt{\frac{\frac{1}{n}\sum _{i=1}^n{\left({\sigma }_s-\overline{{\sigma }_s}\right)}^2}{n-1}}}{\sqrt{\frac{\frac{1}{n}\sum _{i=1}^n{\left({\sigma }_o-\overline{{\sigma }_o}\right)}^2}{n-1}}}}$$

(13)

$$CC={\displaystyle \frac{\sum _{i=1}^n\left[\left({\sigma }_s-\overline{{\sigma }_s}\right)\left({\sigma }_s-\overline{{\sigma }_o}\right)\right]}{\sqrt{\sum _{i=1}^n{\left({\sigma }_s-\overline{{\sigma }_s}\right)}^2{\left({\sigma }_o-\overline{{\sigma }_o}\right)}^2}}}$$

(14)

$$RMSD={\displaystyle \frac{\sqrt{\frac{1}{n}\sum _{i=1}^n{\left({\sigma }_o-{\sigma }_s\right)}^2}}{\sqrt{\frac{\frac{1}{n}\sum _{i=1}^n{\left({\sigma }_o-\overline{{\sigma }_o}\right)}^2}{n-1}}}}$$

(15)

where ${\sigma }_o$ is the observed value at the meteorological station and ${\sigma }_s$ is the simulated value at the GCMs.

2.3.4. Evaluation of POP and GDP Exposures to Compound Events

Here, we further analyze the exposure of the POP and GDP to CDHEs and CHPEs, and the results provide a more intuitive picture of the harmful effects of compound events on human social development. The POP and GDP exposures were obtained by calculating the product of the frequency of the compound event and the values of the POP and GDP indicators [76].

$$EXP=CE\times SEI$$

(16)

where $EXP$ represents the exposure level of the POP/GDP to the CDHEs and CHPEs, while $CE$ represents the frequencies of CDHE and CHPE occurrences, and $SEI$ represents the POP/GDP indicator.

3. Results

3.1. Observed Changes in CDHEs and CHPEs

Figure 3 and Figure 4 show the frequencies and their change rates for the CDHEs and CHPEs in the historical period. The average frequencies of the CDHEs and CHPEs during the historical period (1961–2020) were 0.89 and 0.59 events/year, respectively. The spatial distributions of the CDHEs and CHPEs were complementary, with a higher value in the frequency of CDHEs corresponding to a lower one for the CHPEs. For example, the higher frequencies of CDHEs were mainly located in the PRB and HURB, with the values of 1.47 and 1.35 events/year, respectively. The SWRB had the lowest frequency at 0.3 events/year. The CHPEs were primarily distributed in the eastern HURB and eastern HRB. The HURB had the highest average annual frequency at 1.74 events/year. The SWRB had the lowest value at 0.08 events/year. Overall, both the CDHEs and CHPEs showed significantly increasing trends at rates of 0.11 and 0.08 per decade, respectively, over the historical period (1961–2020). Separately, there was an obvious shift for both compound events. For example, there were decreasing trends at rates of −0.1 and −0.06 events/decade in the first part (1961–1990). However, significantly increasing trends in the CDHEs and CHPEs were found in the second part (1991–2020), both with values of 0.19 events/decade.

The monthly changes from May to September in the frequencies of the CDHEs and CHPEs are shown in Figure 5. Both the CDHEs and CHPEs occurred most frequently in July, with values of 0.32 and 0.31 events/year, followed by August, with annual mean values of 0.25 and 0.18 events. In June, the CDHEs (0.21 events/year) had a higher frequency than the CHPEs (0.09 events/year). The frequencies of the CDHEs and CHPEs for all months showed a non-significantly increasing trend, with the highest rates of 0.05 and 0.04 events per decade in July. Spatially, there was a clear shift in the area of high values of the two compound events between months. The CDHEs roughly experienced a change from southwest–north–south. Higher CDHEs were mainly concentrated in southwestern China in May (Figure 5(a1)) and moved to the north of China in June (Figure 5(a2)), including northwestern, northern, and northeastern China (e.g., IRB, middle and lower YRB, HRB, and SLRB). In July, all but the Tibetan Plateau and southwestern China were characterized by higher values larger than 0.4 events/year (Figure 5(a3)). Subsequently, this shifted to the south of China in August, with the IRB, YZRB, PRB, and SERB being the high-value areas (Figure 5(a4)). There were only a few CDHEs that happened in September (Figure 5(a5)). Similarly, the CHPEs experienced shifting changes between months, showing a southern–northern–southern process. The CHPEs rarely occurred in May and September. In June, the HURB, southwestern IRB, and northeastern HRB were the high-value areas for CHPEs (Figure 5(b2)). In July, the high-value areas of the CHPEs were further expanded, especially in the SLRB, HRB, YRB, and HURB in the north of China (Figure 5(b3)). However, the high-value area in the north shrunk remarkably in August, especially in northern China and northeastern China (Figure 5(b3)).

According to Table 2, we classified the intensity levels of the CDHEs and CHPEs, as shown in Figure 6. Spatially, the CDHEs occurred mainly in the IRB and the south of China, such as in the YZRB and PRB, whereas the CHPEs occurred mainly in the IRB and HURB. Both the CDHEs and CHPEs with the mild level had the highest frequencies during 1961–2020, at 0.34 and 0.25 events/year (Figure 6(a1,b1)). The moderate level had the second-highest values of 0.25 and 0.15 events/year, respectively (Figure 6(a2,b2)). However, the frequencies of the CDHEs and CHPEs were minimal for the extreme level, with about 0.06 and 0.05 events/year, respectively (Figure 6(a4,b4)). Overall, both the CDHEs and CHPEs showed decreasing trends for all four levels in the first half part (1961–1990), with significant decreasing trends being seen in the CDHEs at the severe level (0.03 events per decade) and CHPEs at the mild level (0.02 events per decade) (Figure 6(a7,b5)). However, all the levels of the CDHEs and CHPEs showed significantly increasing trends in the second part (1991–2020), with high rates of 0.07 and 0.05 events per decade for the CDHEs and CHPEs at the severe level (Figure 6(a7,b7)). Therefore, the future risk of compound extreme events may be more serious according to the trends in the second part.

3.2. Projected Variations in CDHEs and CHPEs

We examined the projected changes for the frequencies of two compound events in the three scenarios, as shown in Figure 7. The GCM data show that the projected frequencies for the CDHEs under the SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios were 0.8, 0.94, and 1.23 events/year, respectively, during 2025–2099. The frequency for the SSP1-2.6 scenario will be lower than that of the historical period. Spatially, the projected patterns of the CDHE projections under the future scenarios were similar to those of the historical period; that is, the spatial patterns of the compound extreme events will indeed not change (Figure 7(a1–a3)). From the perspective of the basin scales, the highest frequencies for the CDHEs were found for the HRB (SSP1-2.6) and PRB (SSP2-4.5 and SSP5-8.5), with values of 1.29, 1.43, and 1.76 events/year, respectively. For the three scenarios, the SWRB will suffer the lowest frequencies of CDHEs in the future, with about 0.19, 0.26, and 0.47 events per year, respectively. However, will also show an obvious increase in the frequency of CDHE across Northern China and Northwest China from SSP1-2.6 to SSP5-8.5, such as the IRB, HURB, YRB, and SLRB. Future interannual variations in the CDHEs will be positively correlated with the SSP scenarios, and SSP5-8.5 will have the highest rate of change. The average frequencies of future CHPEs will be 3.92, 5.04, and 7.67 events/year, respectively (Figure 7(b4–b6)). The HRB will have the highest average annual frequency for all three scenarios, 9.85, 11.92, and 15.65 events/year, respectively. The HURB will have the second-highest average frequencies of 8.7, 10.13, and 14.38 events/year, respectively. The spatial distribution of the CHPEs will be mainly concentrated in the north of China, and the IRB, HRB, eastern HURB, and southern SLRB will be the high-value areas of CHPEs. Their areas will expand with the increase in SSP, but the average annual frequency changes in the south of China will be smaller. Similar to the interannual variation in the CDHEs, the SSP5-8.5 scenario will have the highest rate of change. Overall, both the CDHEs and CHPEs will occur frequently in the future but with different magnitudes for various scenarios.

The changes in frequency from May to September for the CDHEs for the three scenarios are shown in Figure 8. The spatial patterns for the CDHEs under the SSP1-2.6 and SSP2-4.5 scenarios will be similar to the historical period, with a major concentration in June–August. However, unlike the historical period, there will be a relatively large increase in the CDHEs in the north of China (e.g., YRB and SLRB). The highest values of annual frequency will be 0.26 (SSP1-2.6), 0.28 (SSP2-4.5), and 0.27 (SSP5-8.5) events/year, respectively. The highest rates of change in the CDHE frequencies will be observed in June, with rates of 0.09 (SSP2-4.5) and 0.13 (SSP5-8.5) events/decade. Under the SSP5-8.5 scenario, the CDHEs will change considerably, where a more even annual frequency will occur between May and September (Figure 8(c1–c5)). The frequencies will increase to 0.2 and 0.21 events/year in May and September, respectively. In May, there will be a higher increase in the north of China, such as the IRB, YRB, HRB, and southern SLRB, whereas it was found in the south of China in September (e.g., the IRB, YZRB, and southwestern HURB). The likely reason for this was that the SSP5-8.5 scenario will have too much precipitation in June, July, and August, which will result in less frequent CDHEs.

Figure 9 shows the spatial changes in the CHPEs for different months under three scenarios. Similarly, future CHPEs will also concentrate in the summer from June to August under the three scenarios, where both the highest frequency and change rate will occur in July, which will be positively correlated with the SSP. The average annual frequencies will be 2.27 (SSP1-2.6), 2.75 (SSP2-4.5), and 3.69 (SSP5-8.5) events/year. The rates of change will be 0.14, 0.35, and 0.58 events/decade, respectively. In terms of the spatial distribution, the CHPEs will mainly be concentrated in the IRB in June (Figure 9(a2,b2,c2)). They will be widely distributed in the north of China in July and August, such as the IRB, HRB, southern SLRB, and eastern HURB (Figure 9(a3,a4,b3,b4,c3,c4)).

The frequencies of occurrence of the four intensity levels of the CDHEs under the future scenarios were analyzed, as shown in Figure 10. The severe level will have the highest frequency for each scenario, with values of 0.29 (SSP1-2.6), 0.37 (SSP2-4.5), and 0.49 (SSP5-8.5) events/year. They will be mainly distributed in most parts of China, except the Tibetan Plateau. As the SSP increases, the annual frequencies of the mild and moderate levels will remain stable, while the frequencies of the severe and extreme levels will increase rapidly. The annual frequencies of the extreme level for the three scenarios will be 0.02, 0.05, and 0.16 events, respectively. The average annual frequency will be second only at the severe level and will occur mainly in the IRB, HRB, YZRB, and PRB. Under the SSP1-2.6 scenario, the four levels of events will tend to increase until 2050 but fluctuate smoothly after 2051 (Figure 10(a5–a8)). Under the SSP2-4.5 and SSP5-8.5 scenarios, there will be weak increasing trends in the four event levels until 2050, but significant increases after 2051, especially for SSP5-8.5 (Figure 10(b5–b8) and (c5–c8)).

Similarly, Figure 11 shows the spatial patterns for the CHPEs in future scenarios. Under the three scenarios, the extreme level will have the highest frequencies, which will be positively correlated with the SSP at 3.12, 4.29, and 6.51 events/year, respectively. They will mainly occur in the IRB, YRB, HURB, HRB, and SLRB in the north of China. The severe level will have the second-highest values of 0.42, 0.4, and 0.43 events/year, respectively. The annual frequencies of the mild and moderate levels will be less in the future. The rate of variation in the annual frequencies of the extreme level will be the highest for the three different scenarios, at 0.26, 0.74, and 1.4 events/decade, respectively (Figure 11(a8–c8)). Under the SSP1-2.6 scenario, the frequency changes of the four levels will be more stable between 2051 and 2099 (Figure 11(a5–a8)), but there will be a significant trend of increasing frequencies of the four levels under the SSP5-8.5 scenario (Figure 11(c5–c8)).

3.3. Future POP and GDP Exposures to Compound Events

The POP and GDP exposures for the CDHEs and CHPEs were analyzed for future scenarios, as shown in Figure 12. It can be seen that the POP exposure for both compound events will be concentrated in the coastal areas of China. The CDHEs will have the highest POP exposure in the SSP2-4.5 scenario at 135.61 person-times/km²·a⁻¹. The HRB will have the highest POP exposure in the SSP2-4.5 scenario at 600.2 person-times/km²·a⁻¹, followed by the HURB and PRB. The CHPEs will have the highest POP exposure under the SSP5-8.5 scenario at 630.1 person-times/km²·a⁻¹. The HRB will have the highest POP exposure under the SSP5-8.5 scenario at 5870.78 person-times/km²·a⁻¹. The spatial distribution of the GDP exposure will be similar to that of the POP exposure. Both the CDHEs and CHPEs will have the highest GDP exposures under the SSP5-8.5 scenario, with 1.37 × 10⁷ USD/km²·a⁻¹ and 6.18 × 10⁷ USD/km²·a⁻¹, respectively. The CDHEs will have the highest GDP exposure in the HURB, with 6.84 × 10⁷ USD/km²·a⁻¹, and the CHPEs will have the highest GDP exposure in the HRB, with 5.33 × 10⁸ USD/km²·a⁻¹. The rates of variation in the POP and GDP exposures for the CDHEs and CHPEs will be positively correlated with the SSP.

4. Discussion

In the context of global warming and accelerated hydrological cycles, our results suggest that the risks of CDHEs and CHPEs will continue to increase regardless of future scenarios. Even in low- and medium-emissions scenarios (SSP1-2.6 and SSP2-4.5), the frequency of CDHEs will remain consistent with the historical period. However, there will be a noticeable increase in the intensity compared with the historical period. Future increases in the CDHE risk have also been shown in the IPCC AR6 report to be associated with an increase in the occurrence of heatwaves. The risk of CDHEs will continue to increase even under a stable future drought scenario [1]. The same will be true for CHPEs, which will lead to a further increase in the risk of future CHPEs due to an increase in the intensity of heatwaves [56]. Thus, our results are also consistent with the IPCC AR6 report and the findings of other scholars. Therefore, to further analyze the relationship between the two extreme events in the compound event, we further calculated the frequencies of CDHEs in each drought event and CHPEs in each heatwave, as shown in Figure 13 and Figure 14. In the three scenarios, the frequency difference of the SSP1-2.6 occurrences in different periods will be the smallest, with 0.08 and 0.01 occurrences, respectively, and the largest difference will be for the SSP5-8.5 scenario, at 0.3 and 0.08 times, respectively. That is, as the SSP increases, the probabilities of CDHEs and CHPEs occurring during each drought/heatwave event will also increase. Sun et al. [56] showed that the intensities of future compound events will be primarily influenced by the intensities of heatwaves. The mechanisms and causes of the increase in compound events are influenced by a combination of aspects. The frequency of compound events in China is also closely linked to atmospheric circulation anomalies, such as the El Niño-Southern Oscillation (ENSO), Atlantic Multidecadal Oscillation (AMO), Indian Ocean Dipole (IOD), and Pacific Decadal Oscillation (PDO) [16,77,78,79,80]. When the ENSO occurs, it leads to the strengthening of subtropical high pressure and western Pacific anticyclones, which provides a favorable environment for the occurrence of heatwaves [77]. This ultimately leads to more CDHEs and CHPEs. Related studies also demonstrated a significantly positive (negative) correlation between dry–hot/hot–wet events in summer in China and the AMO (the East Atlantic pattern, EA) [78], a positive correlation between summer precipitation downstream of the YZRB and the IOD [79], and a close correlation between the PDO and summer precipitation in the HURB [80]. Due to the instability of the atmosphere, the influence of thermal stress under extremely hot weather can lead to the occurrence of hot–wet events [16].

The spatial distribution of socio-economic exposure is closely linked to POP distribution and is mainly affected by a combination of POP change and climate change [81]. As the SSP increases, there will be more and more socio-economic exposure to both CDHEs and CHPEs. The economically developed coastal areas of China will face more severe challenges; not only this, but these high-exposure areas are also the major crop production areas in China, and if these croplands continue to be exposed to CDHEs and CHPEs, this may lead to crop loss or even crop failure.

In this study, the accuracy of the data for projected future changes was crucial as it directly impacted the correctness of the results. The limited spatial resolution of CMIP6 posed challenges for conducting studies at regional scales. Therefore, to ensure the accuracy of the CMIP6 data in this study, we evaluated the 12 CMIP6 models, MME, and bias-corrected MME-Delta against the measured data of the historical period. Classifying 1961–2004 as the calibration period and 2005–2014 as the validation period, the results are shown in Figure 15. Among the three variables, the mean temperature and maximum temperature were simulated better than the precipitation. MME-Delta performed optimally regarding both the RMSD and CC. MME-Delta could compensate well for the bias in CMIP6 regarding precipitation, where the RMSD reached 0.53, which was a reduction of 0.36 compared with MME. NorESM2-LM was better than MME and second only to MME-Delta in the simulation of temperatures. Therefore, the MME-Delta data used in this study were reliable and the method was effective at reducing bias. Although MME-Delta can effectively reduce the errors of CMIP6, CMIP6 still has deficiencies in the simulation of carbon and nitrogen cycling processes, the solution of kinetic equations, and the parameterization of physical processes [82,83,84]. This is an area that still needs to be improved subsequently.

In this study, compound events were identified using a combination of relative and absolute thresholds. The use of this method is justified in China and further improves the accuracy of identifying compound events in this study [49]. Relative thresholds are effective in identifying changes in compound events between different regions with fewer totals [85]. This method enables more accurate and reasonable identification of heatwaves. Drought is influenced by a combination of temperature, precipitation evapotranspiration, etc., and the use of absolute thresholds is more in line with drought classifications [86].

5. Conclusions

This study examined the evolutionary patterns of CDHEs and CHPEs in China by utilizing measured station data for the historical period and CMIP6 data for future scenarios. The spatial and temporal characteristics of their frequency and intensity were analyzed by constructing two magnitude indices, CDHMI and CHPMI, and the changes in future POP and GDP exposures, respectively. The findings were as follows:

(1)

Historically, the frequency of CDHEs was higher than that of CHPEs, with means of 0.89 and 0.59 events/year, respectively. The CDHEs had the highest frequency in the PRB and the CHPEs had the highest one in the HURB. The spatial distributions of the two compound events were complementary. Both compound events occurred most frequently in July.

(2)

The annual mean frequency of the CDHEs will decrease under the SSP1-2.6 scenarios compared with the historical period, but the intensity of the CDHEs will increase. In various scenarios, the occurrence of the CHPEs will be more frequent than that of the CDHEs, with the highest values located in the HRB (CDHEs and CHPEs) and PRB (CDHEs). Under the SSP5-8.5 scenario, there will be a significant increase in May and September compared with the historical period.

(3)

Historically, both the CDHEs and CHPEs had the highest average annual frequencies at the mild level. The CDHEs occurred primarily in the HRB, HURB, western PRB, and northwestern IRB, while the CHPEs occurred primarily in the southern IRB, HRB, and HURB. All levels of the CDHEs and CHPEs were dominated by increasing trends. In the future scenarios, the average annual frequency of the CDHEs at the severe level will be the highest, but it will be the highest for the CHPEs at the extreme level. In the SSP5-8.5 scenario, there will be a significant increase, especially after 2051.

(4)

The POP and GDP exposures of the CDHEs and CHPEs will be primarily concentrated in the coastal regions of China. Under the SSP2-4.5 scenario, the CDHEs will exhibit the highest POP exposure, while the CHPEs under the SSP5-8.5 scenario will display the highest POP exposure.

Due to the high levels of greenhouse gas emissions, the frequency and intensity of compound events have increased with greater magnitudes, particularly in the last 30 years. And their impacts are expected to intensify in the near future. This study can enhance our understanding of compound events across China, provide initial guidance to assess the potential risk of compound events, and aid the evaluation of climate models for future changes in compound events. Moreover, a more comprehensive analysis of the causes of compound events is necessary in the future to enhance their prevention and minimize the impact on human society.